Anthropic Accidentally Deletes Thousands of GitHub Repos While Fixing Source Code Leak

After the source code of Anthropic's Claude Code command-line tool was accidentally leaked, AI enthusiasts uploaded the code to GitHub to study the implementation of its underlying large language model. To stop the spread, Anthropic submitted takedown requests to GitHub under U.S. copyright law.

However, due to a scope judgment error, about 8100 repositories were affected, including branches of the company's own open-source projects, causing user dissatisfaction. Boris Cherny, head of Claude Code, admitted the mistake and has withdrawn most requests, retaining only one repository and its 96 branches.

GitHub has now restored access to most affected repositories. This incident occurs as Anthropic prepares for IPO, potentially impacting its image and legal risks.